NAND is not limited to SSD media: the memory cell architecture resides on circuit boards, which may be housed in SSDs or embedded directly into a server or other device. Still, the majority of NAND flash is delivered via solid state drives in storage arrays, which comprise the core of enterprise nonvolatile flash memory storage.
Nor are flash and SSD interchangeable terms. NAND flash memory is a type of nonvolatile storage, where silicon memory chips persistently store data with or without and external power source. And SSDs are not limited to NAND flash. They also house memory technologies like volatile DRAM.
In the data center, SSDs are an answer to enterprise workloads whose performance suffered with disk-based storage arrays and server storage subsystems. With the growth of hybrid and all-flash arrays, SSD storage serves intensive workloads with very high I/O performance. SSDs have the added advantage of low energy usage, which helps data centers to keep the energy budget items under control.
What is an SSD?
An SSD is a storage device with no moving mechanical parts that houses flash memory and controllers. SSDs use the same external form factors as HDDs, because they are marketed as hard drive replacements. Using the same form factors does not require massive re-engineering of storage arrays at the factory or data center levels.
Since SSDs have no moving parts, they run considerably more quietly, enjoy faster access time, and lower power consumption over hard disk drives. And better reliability developments have made SSDs as durable as disk drives.
How NAND Flash SSDs Work
SSDs store information in memory cell arrays embedded on a circuit board. The memory cells are essentially transistors with floating gates. Each transistor has two gates: one is a source that admits a current, and the other is the drain that expels it. The memory cells act as switches to control the energy flow between the source and drain terminals.
Semiconductors called floating-gate (FG) transistors generate electrical charges to the memory cells, whether connected to an external power source or not (over time a powered-off SSD will leak energy). As long as there is sufficient charge from the FG, the data retains integrity.
Memory cells may house one or more bits per cell. In a single-level cell (SLC), the control gate (CG) will sense if a floating gate is charged with electrons or not. In response, the control gate will record either 0 or 1 bytes. Multi-level cells (MLC) work in a similar way.
The SSD not only houses the interconnected memory cells and circuit boards, but also adds a layer of intelligence with the flash controller.
The speed and performance benefits of SSDs are highly significant.
Advantages of SSD: Why Is a Solid State Drive Better than an HDD?
When comparing SSD vs. HDD, solid state drives truly shine.
- Higher performance. Even the fastest 15K RPM hard drive cannot compete with the performance of NAND flash SSDs. NAND I/O typically achieves 1 Gb/s, while 3D NAND achieves 1.4 GB/s. Newer developments are pushing 3D NAND to 3.0 GB/s. The reason is physics: a hard drive with mechanical components that are in constant usage will break down faster than an SSD that has no mechanical parts. Instead of mechanical arms and read heads, the SSD uses electricity to generate data storage responses. Faster performance means faster boot time, faster data movement, and higher bandwidth.
- Low energy usage. HDDs moving mechanical parts need more energy than the tiny amounts of electrical current shuttling through SSD memory cells. SSDs also avoid the high heat build-up that hundreds of spinning disks generate in a data center, which requires a large investment in HVACs and climate control.
- Commensurate durability. SSD and HDD durability comparisons are more complicated than they might appear. HDD mechanical parts and drive surfaces are more susceptible to environmental damage than SSDs, although new technology is shock-proofing hard drives against physical drops. And SSDs cannot be powered down for long periods of time without a leakage, but powered down HDDs can last decades in environmentally controlled environments. However, SSDs durability is growing thanks to storage intelligence added to the controller. These technologies protect the SSD against data leakage or corruption, and include error-correcting code (ECC), garbage collection, and read and write caching.
Disadvantages of SSDs: Challenges Amid the Speed
Nothing is perfect in the data storage world, and SSDs are no exception. Their disadvantages include higher expense, limited storage capacity, and a shorter delete lifecycle than hard drives.
- Higher cost. SSD dollar-per-GB prices have gone down considerably in the last several years, but so has HDD pricing. Still, flash drive costs have lowered enough so their higher performance becomes cost-effective. Performance is really the key: if HDDs are slowing down transactional databases and other intensive applications, then buying hard drives for affordability is a false economy.
- Lower data storage capacity. NAND SSD capacity lags HDDs thanks to NAND’s memory cell write limitations. The more memory cells on a circuit, the greater density the SSD will achieve. However, planar (2D) NAND can only hold a limited number of memory cells before the cells begin to fail. In response, researchers developed 3D NAND by stacking memory cells vertically as well as horizontally. This enables 3D NAND to achieve higher density, lower power consumption, better endurance, and faster reads/writes, at a lower cost per gigabyte.
- Shorter lifecycle than HDDs. SSDs have a much more limited write cycle than HDDs before failure. The primary reason is that SSDs cannot overwrite existing blocks, but must erase blocks first and then write new data. This process eventually affects the integrity of the memory cell. NAND writes differ according to the number of bits per cell: single-level cell NAND flash supports 50,000 to 100,000 write cycles, multi-level cell generally takes up to 3,000 write cycles, eMLC (enterprise MLC) sustains up to10,000 write cycles, triple-level cells are low at 300-1000 write cycles, and 3D NAND can achieve 1500-3000 write cycles.
- Poor archival media. Businesses want the ability to access, analyze, and monetize their data archives. With their limited number of write cycles, SSDs are not suitable for active archives and repeated analysis on the same data sets. Since the idea of active archives is the ability to access data at will, this overwhelms the number of write cycles that the memory cells can withstand.
What is SSD Good For?
Given these advantages and disadvantages, SSDs are excellent choices for intensive enterprise workloads such as highly transactional databases, web streaming, and dense environments like VDI.
Furthermore, the fast read write speeds of SSD allows them to handle data at the remarkably rapid speeds required of today’s modern business. In fact, businesses can hardly go back to the slower HDD for their top line data usage.
So, is SSD Worth It?
Although the above usage cases are SSD sweet spots in the enterprise, this will raise prices on storage media purchases, and require more SSD swaps than hard drive media. Are SSDs worth the extra time and cost?
In high performance environments, yes. Because SSD form factors are the same as HDDs, replacing disk with SSDs is not a major technology refresh. And because of their higher performance and falling prices, SSDs continue to be a highly competitive storage media in the data center.
SSD Benefits Comparison Chart
Among the benefits of SSD are a much lower failure rate and far faster access time.
|Differentiator||NAND Flash SSD||10K-15K RPM SAS HOD|
HDDs have a higher tolerance for writes, so
|Capacity||In March 2018, Nimbus packed 30 TB into a 2.5″ SSD.||Seagate offered a 16TB hard drive as of Dec. 2018.|
|Access Time||0.1ms||5.5-8.0 ms.
HDD access time is slower because multiple physical operations take time, particularly seek time and rotational latency.
|I/O||2D NAND: 1Gb/s; 3D NAND 1.4 Gb gigabit to /s||It is possible for disk to reach 2D NAND I/0 speeds using clustered high-speed disks.
But this configuration results in underutilized disk, high cost, and complex storage infrastructure along with expensive energy demands.